xref: /openbmc/linux/drivers/crypto/atmel-sha.c (revision 2596e07a)
1 /*
2  * Cryptographic API.
3  *
4  * Support for ATMEL SHA1/SHA256 HW acceleration.
5  *
6  * Copyright (c) 2012 Eukréa Electromatique - ATMEL
7  * Author: Nicolas Royer <nicolas@eukrea.com>
8  *
9  * This program is free software; you can redistribute it and/or modify
10  * it under the terms of the GNU General Public License version 2 as published
11  * by the Free Software Foundation.
12  *
13  * Some ideas are from omap-sham.c drivers.
14  */
15 
16 
17 #include <linux/kernel.h>
18 #include <linux/module.h>
19 #include <linux/slab.h>
20 #include <linux/err.h>
21 #include <linux/clk.h>
22 #include <linux/io.h>
23 #include <linux/hw_random.h>
24 #include <linux/platform_device.h>
25 
26 #include <linux/device.h>
27 #include <linux/init.h>
28 #include <linux/errno.h>
29 #include <linux/interrupt.h>
30 #include <linux/irq.h>
31 #include <linux/scatterlist.h>
32 #include <linux/dma-mapping.h>
33 #include <linux/of_device.h>
34 #include <linux/delay.h>
35 #include <linux/crypto.h>
36 #include <linux/cryptohash.h>
37 #include <crypto/scatterwalk.h>
38 #include <crypto/algapi.h>
39 #include <crypto/sha.h>
40 #include <crypto/hash.h>
41 #include <crypto/internal/hash.h>
42 #include <linux/platform_data/crypto-atmel.h>
43 #include "atmel-sha-regs.h"
44 
45 /* SHA flags */
46 #define SHA_FLAGS_BUSY			BIT(0)
47 #define	SHA_FLAGS_FINAL			BIT(1)
48 #define SHA_FLAGS_DMA_ACTIVE	BIT(2)
49 #define SHA_FLAGS_OUTPUT_READY	BIT(3)
50 #define SHA_FLAGS_INIT			BIT(4)
51 #define SHA_FLAGS_CPU			BIT(5)
52 #define SHA_FLAGS_DMA_READY		BIT(6)
53 
54 #define SHA_FLAGS_FINUP		BIT(16)
55 #define SHA_FLAGS_SG		BIT(17)
56 #define SHA_FLAGS_SHA1		BIT(18)
57 #define SHA_FLAGS_SHA224	BIT(19)
58 #define SHA_FLAGS_SHA256	BIT(20)
59 #define SHA_FLAGS_SHA384	BIT(21)
60 #define SHA_FLAGS_SHA512	BIT(22)
61 #define SHA_FLAGS_ERROR		BIT(23)
62 #define SHA_FLAGS_PAD		BIT(24)
63 
64 #define SHA_OP_UPDATE	1
65 #define SHA_OP_FINAL	2
66 
67 #define SHA_BUFFER_LEN		PAGE_SIZE
68 
69 #define ATMEL_SHA_DMA_THRESHOLD		56
70 
71 struct atmel_sha_caps {
72 	bool	has_dma;
73 	bool	has_dualbuff;
74 	bool	has_sha224;
75 	bool	has_sha_384_512;
76 };
77 
78 struct atmel_sha_dev;
79 
80 struct atmel_sha_reqctx {
81 	struct atmel_sha_dev	*dd;
82 	unsigned long	flags;
83 	unsigned long	op;
84 
85 	u8	digest[SHA512_DIGEST_SIZE] __aligned(sizeof(u32));
86 	u64	digcnt[2];
87 	size_t	bufcnt;
88 	size_t	buflen;
89 	dma_addr_t	dma_addr;
90 
91 	/* walk state */
92 	struct scatterlist	*sg;
93 	unsigned int	offset;	/* offset in current sg */
94 	unsigned int	total;	/* total request */
95 
96 	size_t block_size;
97 
98 	u8	buffer[0] __aligned(sizeof(u32));
99 };
100 
101 struct atmel_sha_ctx {
102 	struct atmel_sha_dev	*dd;
103 
104 	unsigned long		flags;
105 };
106 
107 #define ATMEL_SHA_QUEUE_LENGTH	50
108 
109 struct atmel_sha_dma {
110 	struct dma_chan			*chan;
111 	struct dma_slave_config dma_conf;
112 };
113 
114 struct atmel_sha_dev {
115 	struct list_head	list;
116 	unsigned long		phys_base;
117 	struct device		*dev;
118 	struct clk			*iclk;
119 	int					irq;
120 	void __iomem		*io_base;
121 
122 	spinlock_t		lock;
123 	int			err;
124 	struct tasklet_struct	done_task;
125 
126 	unsigned long		flags;
127 	struct crypto_queue	queue;
128 	struct ahash_request	*req;
129 
130 	struct atmel_sha_dma	dma_lch_in;
131 
132 	struct atmel_sha_caps	caps;
133 
134 	u32	hw_version;
135 };
136 
137 struct atmel_sha_drv {
138 	struct list_head	dev_list;
139 	spinlock_t		lock;
140 };
141 
142 static struct atmel_sha_drv atmel_sha = {
143 	.dev_list = LIST_HEAD_INIT(atmel_sha.dev_list),
144 	.lock = __SPIN_LOCK_UNLOCKED(atmel_sha.lock),
145 };
146 
147 static inline u32 atmel_sha_read(struct atmel_sha_dev *dd, u32 offset)
148 {
149 	return readl_relaxed(dd->io_base + offset);
150 }
151 
152 static inline void atmel_sha_write(struct atmel_sha_dev *dd,
153 					u32 offset, u32 value)
154 {
155 	writel_relaxed(value, dd->io_base + offset);
156 }
157 
158 static size_t atmel_sha_append_sg(struct atmel_sha_reqctx *ctx)
159 {
160 	size_t count;
161 
162 	while ((ctx->bufcnt < ctx->buflen) && ctx->total) {
163 		count = min(ctx->sg->length - ctx->offset, ctx->total);
164 		count = min(count, ctx->buflen - ctx->bufcnt);
165 
166 		if (count <= 0) {
167 			/*
168 			* Check if count <= 0 because the buffer is full or
169 			* because the sg length is 0. In the latest case,
170 			* check if there is another sg in the list, a 0 length
171 			* sg doesn't necessarily mean the end of the sg list.
172 			*/
173 			if ((ctx->sg->length == 0) && !sg_is_last(ctx->sg)) {
174 				ctx->sg = sg_next(ctx->sg);
175 				continue;
176 			} else {
177 				break;
178 			}
179 		}
180 
181 		scatterwalk_map_and_copy(ctx->buffer + ctx->bufcnt, ctx->sg,
182 			ctx->offset, count, 0);
183 
184 		ctx->bufcnt += count;
185 		ctx->offset += count;
186 		ctx->total -= count;
187 
188 		if (ctx->offset == ctx->sg->length) {
189 			ctx->sg = sg_next(ctx->sg);
190 			if (ctx->sg)
191 				ctx->offset = 0;
192 			else
193 				ctx->total = 0;
194 		}
195 	}
196 
197 	return 0;
198 }
199 
200 /*
201  * The purpose of this padding is to ensure that the padded message is a
202  * multiple of 512 bits (SHA1/SHA224/SHA256) or 1024 bits (SHA384/SHA512).
203  * The bit "1" is appended at the end of the message followed by
204  * "padlen-1" zero bits. Then a 64 bits block (SHA1/SHA224/SHA256) or
205  * 128 bits block (SHA384/SHA512) equals to the message length in bits
206  * is appended.
207  *
208  * For SHA1/SHA224/SHA256, padlen is calculated as followed:
209  *  - if message length < 56 bytes then padlen = 56 - message length
210  *  - else padlen = 64 + 56 - message length
211  *
212  * For SHA384/SHA512, padlen is calculated as followed:
213  *  - if message length < 112 bytes then padlen = 112 - message length
214  *  - else padlen = 128 + 112 - message length
215  */
216 static void atmel_sha_fill_padding(struct atmel_sha_reqctx *ctx, int length)
217 {
218 	unsigned int index, padlen;
219 	u64 bits[2];
220 	u64 size[2];
221 
222 	size[0] = ctx->digcnt[0];
223 	size[1] = ctx->digcnt[1];
224 
225 	size[0] += ctx->bufcnt;
226 	if (size[0] < ctx->bufcnt)
227 		size[1]++;
228 
229 	size[0] += length;
230 	if (size[0]  < length)
231 		size[1]++;
232 
233 	bits[1] = cpu_to_be64(size[0] << 3);
234 	bits[0] = cpu_to_be64(size[1] << 3 | size[0] >> 61);
235 
236 	if (ctx->flags & (SHA_FLAGS_SHA384 | SHA_FLAGS_SHA512)) {
237 		index = ctx->bufcnt & 0x7f;
238 		padlen = (index < 112) ? (112 - index) : ((128+112) - index);
239 		*(ctx->buffer + ctx->bufcnt) = 0x80;
240 		memset(ctx->buffer + ctx->bufcnt + 1, 0, padlen-1);
241 		memcpy(ctx->buffer + ctx->bufcnt + padlen, bits, 16);
242 		ctx->bufcnt += padlen + 16;
243 		ctx->flags |= SHA_FLAGS_PAD;
244 	} else {
245 		index = ctx->bufcnt & 0x3f;
246 		padlen = (index < 56) ? (56 - index) : ((64+56) - index);
247 		*(ctx->buffer + ctx->bufcnt) = 0x80;
248 		memset(ctx->buffer + ctx->bufcnt + 1, 0, padlen-1);
249 		memcpy(ctx->buffer + ctx->bufcnt + padlen, &bits[1], 8);
250 		ctx->bufcnt += padlen + 8;
251 		ctx->flags |= SHA_FLAGS_PAD;
252 	}
253 }
254 
255 static int atmel_sha_init(struct ahash_request *req)
256 {
257 	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
258 	struct atmel_sha_ctx *tctx = crypto_ahash_ctx(tfm);
259 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
260 	struct atmel_sha_dev *dd = NULL;
261 	struct atmel_sha_dev *tmp;
262 
263 	spin_lock_bh(&atmel_sha.lock);
264 	if (!tctx->dd) {
265 		list_for_each_entry(tmp, &atmel_sha.dev_list, list) {
266 			dd = tmp;
267 			break;
268 		}
269 		tctx->dd = dd;
270 	} else {
271 		dd = tctx->dd;
272 	}
273 
274 	spin_unlock_bh(&atmel_sha.lock);
275 
276 	ctx->dd = dd;
277 
278 	ctx->flags = 0;
279 
280 	dev_dbg(dd->dev, "init: digest size: %d\n",
281 		crypto_ahash_digestsize(tfm));
282 
283 	switch (crypto_ahash_digestsize(tfm)) {
284 	case SHA1_DIGEST_SIZE:
285 		ctx->flags |= SHA_FLAGS_SHA1;
286 		ctx->block_size = SHA1_BLOCK_SIZE;
287 		break;
288 	case SHA224_DIGEST_SIZE:
289 		ctx->flags |= SHA_FLAGS_SHA224;
290 		ctx->block_size = SHA224_BLOCK_SIZE;
291 		break;
292 	case SHA256_DIGEST_SIZE:
293 		ctx->flags |= SHA_FLAGS_SHA256;
294 		ctx->block_size = SHA256_BLOCK_SIZE;
295 		break;
296 	case SHA384_DIGEST_SIZE:
297 		ctx->flags |= SHA_FLAGS_SHA384;
298 		ctx->block_size = SHA384_BLOCK_SIZE;
299 		break;
300 	case SHA512_DIGEST_SIZE:
301 		ctx->flags |= SHA_FLAGS_SHA512;
302 		ctx->block_size = SHA512_BLOCK_SIZE;
303 		break;
304 	default:
305 		return -EINVAL;
306 		break;
307 	}
308 
309 	ctx->bufcnt = 0;
310 	ctx->digcnt[0] = 0;
311 	ctx->digcnt[1] = 0;
312 	ctx->buflen = SHA_BUFFER_LEN;
313 
314 	return 0;
315 }
316 
317 static void atmel_sha_write_ctrl(struct atmel_sha_dev *dd, int dma)
318 {
319 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
320 	u32 valcr = 0, valmr = SHA_MR_MODE_AUTO;
321 
322 	if (likely(dma)) {
323 		if (!dd->caps.has_dma)
324 			atmel_sha_write(dd, SHA_IER, SHA_INT_TXBUFE);
325 		valmr = SHA_MR_MODE_PDC;
326 		if (dd->caps.has_dualbuff)
327 			valmr |= SHA_MR_DUALBUFF;
328 	} else {
329 		atmel_sha_write(dd, SHA_IER, SHA_INT_DATARDY);
330 	}
331 
332 	if (ctx->flags & SHA_FLAGS_SHA1)
333 		valmr |= SHA_MR_ALGO_SHA1;
334 	else if (ctx->flags & SHA_FLAGS_SHA224)
335 		valmr |= SHA_MR_ALGO_SHA224;
336 	else if (ctx->flags & SHA_FLAGS_SHA256)
337 		valmr |= SHA_MR_ALGO_SHA256;
338 	else if (ctx->flags & SHA_FLAGS_SHA384)
339 		valmr |= SHA_MR_ALGO_SHA384;
340 	else if (ctx->flags & SHA_FLAGS_SHA512)
341 		valmr |= SHA_MR_ALGO_SHA512;
342 
343 	/* Setting CR_FIRST only for the first iteration */
344 	if (!(ctx->digcnt[0] || ctx->digcnt[1]))
345 		valcr = SHA_CR_FIRST;
346 
347 	atmel_sha_write(dd, SHA_CR, valcr);
348 	atmel_sha_write(dd, SHA_MR, valmr);
349 }
350 
351 static int atmel_sha_xmit_cpu(struct atmel_sha_dev *dd, const u8 *buf,
352 			      size_t length, int final)
353 {
354 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
355 	int count, len32;
356 	const u32 *buffer = (const u32 *)buf;
357 
358 	dev_dbg(dd->dev, "xmit_cpu: digcnt: 0x%llx 0x%llx, length: %d, final: %d\n",
359 		ctx->digcnt[1], ctx->digcnt[0], length, final);
360 
361 	atmel_sha_write_ctrl(dd, 0);
362 
363 	/* should be non-zero before next lines to disable clocks later */
364 	ctx->digcnt[0] += length;
365 	if (ctx->digcnt[0] < length)
366 		ctx->digcnt[1]++;
367 
368 	if (final)
369 		dd->flags |= SHA_FLAGS_FINAL; /* catch last interrupt */
370 
371 	len32 = DIV_ROUND_UP(length, sizeof(u32));
372 
373 	dd->flags |= SHA_FLAGS_CPU;
374 
375 	for (count = 0; count < len32; count++)
376 		atmel_sha_write(dd, SHA_REG_DIN(count), buffer[count]);
377 
378 	return -EINPROGRESS;
379 }
380 
381 static int atmel_sha_xmit_pdc(struct atmel_sha_dev *dd, dma_addr_t dma_addr1,
382 		size_t length1, dma_addr_t dma_addr2, size_t length2, int final)
383 {
384 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
385 	int len32;
386 
387 	dev_dbg(dd->dev, "xmit_pdc: digcnt: 0x%llx 0x%llx, length: %d, final: %d\n",
388 		ctx->digcnt[1], ctx->digcnt[0], length1, final);
389 
390 	len32 = DIV_ROUND_UP(length1, sizeof(u32));
391 	atmel_sha_write(dd, SHA_PTCR, SHA_PTCR_TXTDIS);
392 	atmel_sha_write(dd, SHA_TPR, dma_addr1);
393 	atmel_sha_write(dd, SHA_TCR, len32);
394 
395 	len32 = DIV_ROUND_UP(length2, sizeof(u32));
396 	atmel_sha_write(dd, SHA_TNPR, dma_addr2);
397 	atmel_sha_write(dd, SHA_TNCR, len32);
398 
399 	atmel_sha_write_ctrl(dd, 1);
400 
401 	/* should be non-zero before next lines to disable clocks later */
402 	ctx->digcnt[0] += length1;
403 	if (ctx->digcnt[0] < length1)
404 		ctx->digcnt[1]++;
405 
406 	if (final)
407 		dd->flags |= SHA_FLAGS_FINAL; /* catch last interrupt */
408 
409 	dd->flags |=  SHA_FLAGS_DMA_ACTIVE;
410 
411 	/* Start DMA transfer */
412 	atmel_sha_write(dd, SHA_PTCR, SHA_PTCR_TXTEN);
413 
414 	return -EINPROGRESS;
415 }
416 
417 static void atmel_sha_dma_callback(void *data)
418 {
419 	struct atmel_sha_dev *dd = data;
420 
421 	/* dma_lch_in - completed - wait DATRDY */
422 	atmel_sha_write(dd, SHA_IER, SHA_INT_DATARDY);
423 }
424 
425 static int atmel_sha_xmit_dma(struct atmel_sha_dev *dd, dma_addr_t dma_addr1,
426 		size_t length1, dma_addr_t dma_addr2, size_t length2, int final)
427 {
428 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
429 	struct dma_async_tx_descriptor	*in_desc;
430 	struct scatterlist sg[2];
431 
432 	dev_dbg(dd->dev, "xmit_dma: digcnt: 0x%llx 0x%llx, length: %d, final: %d\n",
433 		ctx->digcnt[1], ctx->digcnt[0], length1, final);
434 
435 	dd->dma_lch_in.dma_conf.src_maxburst = 16;
436 	dd->dma_lch_in.dma_conf.dst_maxburst = 16;
437 
438 	dmaengine_slave_config(dd->dma_lch_in.chan, &dd->dma_lch_in.dma_conf);
439 
440 	if (length2) {
441 		sg_init_table(sg, 2);
442 		sg_dma_address(&sg[0]) = dma_addr1;
443 		sg_dma_len(&sg[0]) = length1;
444 		sg_dma_address(&sg[1]) = dma_addr2;
445 		sg_dma_len(&sg[1]) = length2;
446 		in_desc = dmaengine_prep_slave_sg(dd->dma_lch_in.chan, sg, 2,
447 			DMA_MEM_TO_DEV, DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
448 	} else {
449 		sg_init_table(sg, 1);
450 		sg_dma_address(&sg[0]) = dma_addr1;
451 		sg_dma_len(&sg[0]) = length1;
452 		in_desc = dmaengine_prep_slave_sg(dd->dma_lch_in.chan, sg, 1,
453 			DMA_MEM_TO_DEV, DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
454 	}
455 	if (!in_desc)
456 		return -EINVAL;
457 
458 	in_desc->callback = atmel_sha_dma_callback;
459 	in_desc->callback_param = dd;
460 
461 	atmel_sha_write_ctrl(dd, 1);
462 
463 	/* should be non-zero before next lines to disable clocks later */
464 	ctx->digcnt[0] += length1;
465 	if (ctx->digcnt[0] < length1)
466 		ctx->digcnt[1]++;
467 
468 	if (final)
469 		dd->flags |= SHA_FLAGS_FINAL; /* catch last interrupt */
470 
471 	dd->flags |=  SHA_FLAGS_DMA_ACTIVE;
472 
473 	/* Start DMA transfer */
474 	dmaengine_submit(in_desc);
475 	dma_async_issue_pending(dd->dma_lch_in.chan);
476 
477 	return -EINPROGRESS;
478 }
479 
480 static int atmel_sha_xmit_start(struct atmel_sha_dev *dd, dma_addr_t dma_addr1,
481 		size_t length1, dma_addr_t dma_addr2, size_t length2, int final)
482 {
483 	if (dd->caps.has_dma)
484 		return atmel_sha_xmit_dma(dd, dma_addr1, length1,
485 				dma_addr2, length2, final);
486 	else
487 		return atmel_sha_xmit_pdc(dd, dma_addr1, length1,
488 				dma_addr2, length2, final);
489 }
490 
491 static int atmel_sha_update_cpu(struct atmel_sha_dev *dd)
492 {
493 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
494 	int bufcnt;
495 
496 	atmel_sha_append_sg(ctx);
497 	atmel_sha_fill_padding(ctx, 0);
498 	bufcnt = ctx->bufcnt;
499 	ctx->bufcnt = 0;
500 
501 	return atmel_sha_xmit_cpu(dd, ctx->buffer, bufcnt, 1);
502 }
503 
504 static int atmel_sha_xmit_dma_map(struct atmel_sha_dev *dd,
505 					struct atmel_sha_reqctx *ctx,
506 					size_t length, int final)
507 {
508 	ctx->dma_addr = dma_map_single(dd->dev, ctx->buffer,
509 				ctx->buflen + ctx->block_size, DMA_TO_DEVICE);
510 	if (dma_mapping_error(dd->dev, ctx->dma_addr)) {
511 		dev_err(dd->dev, "dma %u bytes error\n", ctx->buflen +
512 				ctx->block_size);
513 		return -EINVAL;
514 	}
515 
516 	ctx->flags &= ~SHA_FLAGS_SG;
517 
518 	/* next call does not fail... so no unmap in the case of error */
519 	return atmel_sha_xmit_start(dd, ctx->dma_addr, length, 0, 0, final);
520 }
521 
522 static int atmel_sha_update_dma_slow(struct atmel_sha_dev *dd)
523 {
524 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
525 	unsigned int final;
526 	size_t count;
527 
528 	atmel_sha_append_sg(ctx);
529 
530 	final = (ctx->flags & SHA_FLAGS_FINUP) && !ctx->total;
531 
532 	dev_dbg(dd->dev, "slow: bufcnt: %u, digcnt: 0x%llx 0x%llx, final: %d\n",
533 		 ctx->bufcnt, ctx->digcnt[1], ctx->digcnt[0], final);
534 
535 	if (final)
536 		atmel_sha_fill_padding(ctx, 0);
537 
538 	if (final || (ctx->bufcnt == ctx->buflen)) {
539 		count = ctx->bufcnt;
540 		ctx->bufcnt = 0;
541 		return atmel_sha_xmit_dma_map(dd, ctx, count, final);
542 	}
543 
544 	return 0;
545 }
546 
547 static int atmel_sha_update_dma_start(struct atmel_sha_dev *dd)
548 {
549 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
550 	unsigned int length, final, tail;
551 	struct scatterlist *sg;
552 	unsigned int count;
553 
554 	if (!ctx->total)
555 		return 0;
556 
557 	if (ctx->bufcnt || ctx->offset)
558 		return atmel_sha_update_dma_slow(dd);
559 
560 	dev_dbg(dd->dev, "fast: digcnt: 0x%llx 0x%llx, bufcnt: %u, total: %u\n",
561 		ctx->digcnt[1], ctx->digcnt[0], ctx->bufcnt, ctx->total);
562 
563 	sg = ctx->sg;
564 
565 	if (!IS_ALIGNED(sg->offset, sizeof(u32)))
566 		return atmel_sha_update_dma_slow(dd);
567 
568 	if (!sg_is_last(sg) && !IS_ALIGNED(sg->length, ctx->block_size))
569 		/* size is not ctx->block_size aligned */
570 		return atmel_sha_update_dma_slow(dd);
571 
572 	length = min(ctx->total, sg->length);
573 
574 	if (sg_is_last(sg)) {
575 		if (!(ctx->flags & SHA_FLAGS_FINUP)) {
576 			/* not last sg must be ctx->block_size aligned */
577 			tail = length & (ctx->block_size - 1);
578 			length -= tail;
579 		}
580 	}
581 
582 	ctx->total -= length;
583 	ctx->offset = length; /* offset where to start slow */
584 
585 	final = (ctx->flags & SHA_FLAGS_FINUP) && !ctx->total;
586 
587 	/* Add padding */
588 	if (final) {
589 		tail = length & (ctx->block_size - 1);
590 		length -= tail;
591 		ctx->total += tail;
592 		ctx->offset = length; /* offset where to start slow */
593 
594 		sg = ctx->sg;
595 		atmel_sha_append_sg(ctx);
596 
597 		atmel_sha_fill_padding(ctx, length);
598 
599 		ctx->dma_addr = dma_map_single(dd->dev, ctx->buffer,
600 			ctx->buflen + ctx->block_size, DMA_TO_DEVICE);
601 		if (dma_mapping_error(dd->dev, ctx->dma_addr)) {
602 			dev_err(dd->dev, "dma %u bytes error\n",
603 				ctx->buflen + ctx->block_size);
604 			return -EINVAL;
605 		}
606 
607 		if (length == 0) {
608 			ctx->flags &= ~SHA_FLAGS_SG;
609 			count = ctx->bufcnt;
610 			ctx->bufcnt = 0;
611 			return atmel_sha_xmit_start(dd, ctx->dma_addr, count, 0,
612 					0, final);
613 		} else {
614 			ctx->sg = sg;
615 			if (!dma_map_sg(dd->dev, ctx->sg, 1,
616 				DMA_TO_DEVICE)) {
617 					dev_err(dd->dev, "dma_map_sg  error\n");
618 					return -EINVAL;
619 			}
620 
621 			ctx->flags |= SHA_FLAGS_SG;
622 
623 			count = ctx->bufcnt;
624 			ctx->bufcnt = 0;
625 			return atmel_sha_xmit_start(dd, sg_dma_address(ctx->sg),
626 					length, ctx->dma_addr, count, final);
627 		}
628 	}
629 
630 	if (!dma_map_sg(dd->dev, ctx->sg, 1, DMA_TO_DEVICE)) {
631 		dev_err(dd->dev, "dma_map_sg  error\n");
632 		return -EINVAL;
633 	}
634 
635 	ctx->flags |= SHA_FLAGS_SG;
636 
637 	/* next call does not fail... so no unmap in the case of error */
638 	return atmel_sha_xmit_start(dd, sg_dma_address(ctx->sg), length, 0,
639 								0, final);
640 }
641 
642 static int atmel_sha_update_dma_stop(struct atmel_sha_dev *dd)
643 {
644 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
645 
646 	if (ctx->flags & SHA_FLAGS_SG) {
647 		dma_unmap_sg(dd->dev, ctx->sg, 1, DMA_TO_DEVICE);
648 		if (ctx->sg->length == ctx->offset) {
649 			ctx->sg = sg_next(ctx->sg);
650 			if (ctx->sg)
651 				ctx->offset = 0;
652 		}
653 		if (ctx->flags & SHA_FLAGS_PAD) {
654 			dma_unmap_single(dd->dev, ctx->dma_addr,
655 				ctx->buflen + ctx->block_size, DMA_TO_DEVICE);
656 		}
657 	} else {
658 		dma_unmap_single(dd->dev, ctx->dma_addr, ctx->buflen +
659 						ctx->block_size, DMA_TO_DEVICE);
660 	}
661 
662 	return 0;
663 }
664 
665 static int atmel_sha_update_req(struct atmel_sha_dev *dd)
666 {
667 	struct ahash_request *req = dd->req;
668 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
669 	int err;
670 
671 	dev_dbg(dd->dev, "update_req: total: %u, digcnt: 0x%llx 0x%llx\n",
672 		ctx->total, ctx->digcnt[1], ctx->digcnt[0]);
673 
674 	if (ctx->flags & SHA_FLAGS_CPU)
675 		err = atmel_sha_update_cpu(dd);
676 	else
677 		err = atmel_sha_update_dma_start(dd);
678 
679 	/* wait for dma completion before can take more data */
680 	dev_dbg(dd->dev, "update: err: %d, digcnt: 0x%llx 0%llx\n",
681 			err, ctx->digcnt[1], ctx->digcnt[0]);
682 
683 	return err;
684 }
685 
686 static int atmel_sha_final_req(struct atmel_sha_dev *dd)
687 {
688 	struct ahash_request *req = dd->req;
689 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
690 	int err = 0;
691 	int count;
692 
693 	if (ctx->bufcnt >= ATMEL_SHA_DMA_THRESHOLD) {
694 		atmel_sha_fill_padding(ctx, 0);
695 		count = ctx->bufcnt;
696 		ctx->bufcnt = 0;
697 		err = atmel_sha_xmit_dma_map(dd, ctx, count, 1);
698 	}
699 	/* faster to handle last block with cpu */
700 	else {
701 		atmel_sha_fill_padding(ctx, 0);
702 		count = ctx->bufcnt;
703 		ctx->bufcnt = 0;
704 		err = atmel_sha_xmit_cpu(dd, ctx->buffer, count, 1);
705 	}
706 
707 	dev_dbg(dd->dev, "final_req: err: %d\n", err);
708 
709 	return err;
710 }
711 
712 static void atmel_sha_copy_hash(struct ahash_request *req)
713 {
714 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
715 	u32 *hash = (u32 *)ctx->digest;
716 	int i;
717 
718 	if (ctx->flags & SHA_FLAGS_SHA1)
719 		for (i = 0; i < SHA1_DIGEST_SIZE / sizeof(u32); i++)
720 			hash[i] = atmel_sha_read(ctx->dd, SHA_REG_DIGEST(i));
721 	else if (ctx->flags & SHA_FLAGS_SHA224)
722 		for (i = 0; i < SHA224_DIGEST_SIZE / sizeof(u32); i++)
723 			hash[i] = atmel_sha_read(ctx->dd, SHA_REG_DIGEST(i));
724 	else if (ctx->flags & SHA_FLAGS_SHA256)
725 		for (i = 0; i < SHA256_DIGEST_SIZE / sizeof(u32); i++)
726 			hash[i] = atmel_sha_read(ctx->dd, SHA_REG_DIGEST(i));
727 	else if (ctx->flags & SHA_FLAGS_SHA384)
728 		for (i = 0; i < SHA384_DIGEST_SIZE / sizeof(u32); i++)
729 			hash[i] = atmel_sha_read(ctx->dd, SHA_REG_DIGEST(i));
730 	else
731 		for (i = 0; i < SHA512_DIGEST_SIZE / sizeof(u32); i++)
732 			hash[i] = atmel_sha_read(ctx->dd, SHA_REG_DIGEST(i));
733 }
734 
735 static void atmel_sha_copy_ready_hash(struct ahash_request *req)
736 {
737 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
738 
739 	if (!req->result)
740 		return;
741 
742 	if (ctx->flags & SHA_FLAGS_SHA1)
743 		memcpy(req->result, ctx->digest, SHA1_DIGEST_SIZE);
744 	else if (ctx->flags & SHA_FLAGS_SHA224)
745 		memcpy(req->result, ctx->digest, SHA224_DIGEST_SIZE);
746 	else if (ctx->flags & SHA_FLAGS_SHA256)
747 		memcpy(req->result, ctx->digest, SHA256_DIGEST_SIZE);
748 	else if (ctx->flags & SHA_FLAGS_SHA384)
749 		memcpy(req->result, ctx->digest, SHA384_DIGEST_SIZE);
750 	else
751 		memcpy(req->result, ctx->digest, SHA512_DIGEST_SIZE);
752 }
753 
754 static int atmel_sha_finish(struct ahash_request *req)
755 {
756 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
757 	struct atmel_sha_dev *dd = ctx->dd;
758 
759 	if (ctx->digcnt[0] || ctx->digcnt[1])
760 		atmel_sha_copy_ready_hash(req);
761 
762 	dev_dbg(dd->dev, "digcnt: 0x%llx 0x%llx, bufcnt: %d\n", ctx->digcnt[1],
763 		ctx->digcnt[0], ctx->bufcnt);
764 
765 	return 0;
766 }
767 
768 static void atmel_sha_finish_req(struct ahash_request *req, int err)
769 {
770 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
771 	struct atmel_sha_dev *dd = ctx->dd;
772 
773 	if (!err) {
774 		atmel_sha_copy_hash(req);
775 		if (SHA_FLAGS_FINAL & dd->flags)
776 			err = atmel_sha_finish(req);
777 	} else {
778 		ctx->flags |= SHA_FLAGS_ERROR;
779 	}
780 
781 	/* atomic operation is not needed here */
782 	dd->flags &= ~(SHA_FLAGS_BUSY | SHA_FLAGS_FINAL | SHA_FLAGS_CPU |
783 			SHA_FLAGS_DMA_READY | SHA_FLAGS_OUTPUT_READY);
784 
785 	clk_disable(dd->iclk);
786 
787 	if (req->base.complete)
788 		req->base.complete(&req->base, err);
789 
790 	/* handle new request */
791 	tasklet_schedule(&dd->done_task);
792 }
793 
794 static int atmel_sha_hw_init(struct atmel_sha_dev *dd)
795 {
796 	int err;
797 
798 	err = clk_enable(dd->iclk);
799 	if (err)
800 		return err;
801 
802 	if (!(SHA_FLAGS_INIT & dd->flags)) {
803 		atmel_sha_write(dd, SHA_CR, SHA_CR_SWRST);
804 		dd->flags |= SHA_FLAGS_INIT;
805 		dd->err = 0;
806 	}
807 
808 	return 0;
809 }
810 
811 static inline unsigned int atmel_sha_get_version(struct atmel_sha_dev *dd)
812 {
813 	return atmel_sha_read(dd, SHA_HW_VERSION) & 0x00000fff;
814 }
815 
816 static void atmel_sha_hw_version_init(struct atmel_sha_dev *dd)
817 {
818 	atmel_sha_hw_init(dd);
819 
820 	dd->hw_version = atmel_sha_get_version(dd);
821 
822 	dev_info(dd->dev,
823 			"version: 0x%x\n", dd->hw_version);
824 
825 	clk_disable(dd->iclk);
826 }
827 
828 static int atmel_sha_handle_queue(struct atmel_sha_dev *dd,
829 				  struct ahash_request *req)
830 {
831 	struct crypto_async_request *async_req, *backlog;
832 	struct atmel_sha_reqctx *ctx;
833 	unsigned long flags;
834 	int err = 0, ret = 0;
835 
836 	spin_lock_irqsave(&dd->lock, flags);
837 	if (req)
838 		ret = ahash_enqueue_request(&dd->queue, req);
839 
840 	if (SHA_FLAGS_BUSY & dd->flags) {
841 		spin_unlock_irqrestore(&dd->lock, flags);
842 		return ret;
843 	}
844 
845 	backlog = crypto_get_backlog(&dd->queue);
846 	async_req = crypto_dequeue_request(&dd->queue);
847 	if (async_req)
848 		dd->flags |= SHA_FLAGS_BUSY;
849 
850 	spin_unlock_irqrestore(&dd->lock, flags);
851 
852 	if (!async_req)
853 		return ret;
854 
855 	if (backlog)
856 		backlog->complete(backlog, -EINPROGRESS);
857 
858 	req = ahash_request_cast(async_req);
859 	dd->req = req;
860 	ctx = ahash_request_ctx(req);
861 
862 	dev_dbg(dd->dev, "handling new req, op: %lu, nbytes: %d\n",
863 						ctx->op, req->nbytes);
864 
865 	err = atmel_sha_hw_init(dd);
866 
867 	if (err)
868 		goto err1;
869 
870 	if (ctx->op == SHA_OP_UPDATE) {
871 		err = atmel_sha_update_req(dd);
872 		if (err != -EINPROGRESS && (ctx->flags & SHA_FLAGS_FINUP))
873 			/* no final() after finup() */
874 			err = atmel_sha_final_req(dd);
875 	} else if (ctx->op == SHA_OP_FINAL) {
876 		err = atmel_sha_final_req(dd);
877 	}
878 
879 err1:
880 	if (err != -EINPROGRESS)
881 		/* done_task will not finish it, so do it here */
882 		atmel_sha_finish_req(req, err);
883 
884 	dev_dbg(dd->dev, "exit, err: %d\n", err);
885 
886 	return ret;
887 }
888 
889 static int atmel_sha_enqueue(struct ahash_request *req, unsigned int op)
890 {
891 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
892 	struct atmel_sha_ctx *tctx = crypto_tfm_ctx(req->base.tfm);
893 	struct atmel_sha_dev *dd = tctx->dd;
894 
895 	ctx->op = op;
896 
897 	return atmel_sha_handle_queue(dd, req);
898 }
899 
900 static int atmel_sha_update(struct ahash_request *req)
901 {
902 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
903 
904 	if (!req->nbytes)
905 		return 0;
906 
907 	ctx->total = req->nbytes;
908 	ctx->sg = req->src;
909 	ctx->offset = 0;
910 
911 	if (ctx->flags & SHA_FLAGS_FINUP) {
912 		if (ctx->bufcnt + ctx->total < ATMEL_SHA_DMA_THRESHOLD)
913 			/* faster to use CPU for short transfers */
914 			ctx->flags |= SHA_FLAGS_CPU;
915 	} else if (ctx->bufcnt + ctx->total < ctx->buflen) {
916 		atmel_sha_append_sg(ctx);
917 		return 0;
918 	}
919 	return atmel_sha_enqueue(req, SHA_OP_UPDATE);
920 }
921 
922 static int atmel_sha_final(struct ahash_request *req)
923 {
924 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
925 	struct atmel_sha_ctx *tctx = crypto_tfm_ctx(req->base.tfm);
926 	struct atmel_sha_dev *dd = tctx->dd;
927 
928 	int err = 0;
929 
930 	ctx->flags |= SHA_FLAGS_FINUP;
931 
932 	if (ctx->flags & SHA_FLAGS_ERROR)
933 		return 0; /* uncompleted hash is not needed */
934 
935 	if (ctx->bufcnt) {
936 		return atmel_sha_enqueue(req, SHA_OP_FINAL);
937 	} else if (!(ctx->flags & SHA_FLAGS_PAD)) { /* add padding */
938 		err = atmel_sha_hw_init(dd);
939 		if (err)
940 			goto err1;
941 
942 		dd->flags |= SHA_FLAGS_BUSY;
943 		err = atmel_sha_final_req(dd);
944 	} else {
945 		/* copy ready hash (+ finalize hmac) */
946 		return atmel_sha_finish(req);
947 	}
948 
949 err1:
950 	if (err != -EINPROGRESS)
951 		/* done_task will not finish it, so do it here */
952 		atmel_sha_finish_req(req, err);
953 
954 	return err;
955 }
956 
957 static int atmel_sha_finup(struct ahash_request *req)
958 {
959 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
960 	int err1, err2;
961 
962 	ctx->flags |= SHA_FLAGS_FINUP;
963 
964 	err1 = atmel_sha_update(req);
965 	if (err1 == -EINPROGRESS || err1 == -EBUSY)
966 		return err1;
967 
968 	/*
969 	 * final() has to be always called to cleanup resources
970 	 * even if udpate() failed, except EINPROGRESS
971 	 */
972 	err2 = atmel_sha_final(req);
973 
974 	return err1 ?: err2;
975 }
976 
977 static int atmel_sha_digest(struct ahash_request *req)
978 {
979 	return atmel_sha_init(req) ?: atmel_sha_finup(req);
980 }
981 
982 static int atmel_sha_cra_init(struct crypto_tfm *tfm)
983 {
984 	crypto_ahash_set_reqsize(__crypto_ahash_cast(tfm),
985 				 sizeof(struct atmel_sha_reqctx) +
986 				 SHA_BUFFER_LEN + SHA512_BLOCK_SIZE);
987 
988 	return 0;
989 }
990 
991 static struct ahash_alg sha_1_256_algs[] = {
992 {
993 	.init		= atmel_sha_init,
994 	.update		= atmel_sha_update,
995 	.final		= atmel_sha_final,
996 	.finup		= atmel_sha_finup,
997 	.digest		= atmel_sha_digest,
998 	.halg = {
999 		.digestsize	= SHA1_DIGEST_SIZE,
1000 		.base	= {
1001 			.cra_name		= "sha1",
1002 			.cra_driver_name	= "atmel-sha1",
1003 			.cra_priority		= 100,
1004 			.cra_flags		= CRYPTO_ALG_ASYNC,
1005 			.cra_blocksize		= SHA1_BLOCK_SIZE,
1006 			.cra_ctxsize		= sizeof(struct atmel_sha_ctx),
1007 			.cra_alignmask		= 0,
1008 			.cra_module		= THIS_MODULE,
1009 			.cra_init		= atmel_sha_cra_init,
1010 		}
1011 	}
1012 },
1013 {
1014 	.init		= atmel_sha_init,
1015 	.update		= atmel_sha_update,
1016 	.final		= atmel_sha_final,
1017 	.finup		= atmel_sha_finup,
1018 	.digest		= atmel_sha_digest,
1019 	.halg = {
1020 		.digestsize	= SHA256_DIGEST_SIZE,
1021 		.base	= {
1022 			.cra_name		= "sha256",
1023 			.cra_driver_name	= "atmel-sha256",
1024 			.cra_priority		= 100,
1025 			.cra_flags		= CRYPTO_ALG_ASYNC,
1026 			.cra_blocksize		= SHA256_BLOCK_SIZE,
1027 			.cra_ctxsize		= sizeof(struct atmel_sha_ctx),
1028 			.cra_alignmask		= 0,
1029 			.cra_module		= THIS_MODULE,
1030 			.cra_init		= atmel_sha_cra_init,
1031 		}
1032 	}
1033 },
1034 };
1035 
1036 static struct ahash_alg sha_224_alg = {
1037 	.init		= atmel_sha_init,
1038 	.update		= atmel_sha_update,
1039 	.final		= atmel_sha_final,
1040 	.finup		= atmel_sha_finup,
1041 	.digest		= atmel_sha_digest,
1042 	.halg = {
1043 		.digestsize	= SHA224_DIGEST_SIZE,
1044 		.base	= {
1045 			.cra_name		= "sha224",
1046 			.cra_driver_name	= "atmel-sha224",
1047 			.cra_priority		= 100,
1048 			.cra_flags		= CRYPTO_ALG_ASYNC,
1049 			.cra_blocksize		= SHA224_BLOCK_SIZE,
1050 			.cra_ctxsize		= sizeof(struct atmel_sha_ctx),
1051 			.cra_alignmask		= 0,
1052 			.cra_module		= THIS_MODULE,
1053 			.cra_init		= atmel_sha_cra_init,
1054 		}
1055 	}
1056 };
1057 
1058 static struct ahash_alg sha_384_512_algs[] = {
1059 {
1060 	.init		= atmel_sha_init,
1061 	.update		= atmel_sha_update,
1062 	.final		= atmel_sha_final,
1063 	.finup		= atmel_sha_finup,
1064 	.digest		= atmel_sha_digest,
1065 	.halg = {
1066 		.digestsize	= SHA384_DIGEST_SIZE,
1067 		.base	= {
1068 			.cra_name		= "sha384",
1069 			.cra_driver_name	= "atmel-sha384",
1070 			.cra_priority		= 100,
1071 			.cra_flags		= CRYPTO_ALG_ASYNC,
1072 			.cra_blocksize		= SHA384_BLOCK_SIZE,
1073 			.cra_ctxsize		= sizeof(struct atmel_sha_ctx),
1074 			.cra_alignmask		= 0x3,
1075 			.cra_module		= THIS_MODULE,
1076 			.cra_init		= atmel_sha_cra_init,
1077 		}
1078 	}
1079 },
1080 {
1081 	.init		= atmel_sha_init,
1082 	.update		= atmel_sha_update,
1083 	.final		= atmel_sha_final,
1084 	.finup		= atmel_sha_finup,
1085 	.digest		= atmel_sha_digest,
1086 	.halg = {
1087 		.digestsize	= SHA512_DIGEST_SIZE,
1088 		.base	= {
1089 			.cra_name		= "sha512",
1090 			.cra_driver_name	= "atmel-sha512",
1091 			.cra_priority		= 100,
1092 			.cra_flags		= CRYPTO_ALG_ASYNC,
1093 			.cra_blocksize		= SHA512_BLOCK_SIZE,
1094 			.cra_ctxsize		= sizeof(struct atmel_sha_ctx),
1095 			.cra_alignmask		= 0x3,
1096 			.cra_module		= THIS_MODULE,
1097 			.cra_init		= atmel_sha_cra_init,
1098 		}
1099 	}
1100 },
1101 };
1102 
1103 static void atmel_sha_done_task(unsigned long data)
1104 {
1105 	struct atmel_sha_dev *dd = (struct atmel_sha_dev *)data;
1106 	int err = 0;
1107 
1108 	if (!(SHA_FLAGS_BUSY & dd->flags)) {
1109 		atmel_sha_handle_queue(dd, NULL);
1110 		return;
1111 	}
1112 
1113 	if (SHA_FLAGS_CPU & dd->flags) {
1114 		if (SHA_FLAGS_OUTPUT_READY & dd->flags) {
1115 			dd->flags &= ~SHA_FLAGS_OUTPUT_READY;
1116 			goto finish;
1117 		}
1118 	} else if (SHA_FLAGS_DMA_READY & dd->flags) {
1119 		if (SHA_FLAGS_DMA_ACTIVE & dd->flags) {
1120 			dd->flags &= ~SHA_FLAGS_DMA_ACTIVE;
1121 			atmel_sha_update_dma_stop(dd);
1122 			if (dd->err) {
1123 				err = dd->err;
1124 				goto finish;
1125 			}
1126 		}
1127 		if (SHA_FLAGS_OUTPUT_READY & dd->flags) {
1128 			/* hash or semi-hash ready */
1129 			dd->flags &= ~(SHA_FLAGS_DMA_READY |
1130 						SHA_FLAGS_OUTPUT_READY);
1131 			err = atmel_sha_update_dma_start(dd);
1132 			if (err != -EINPROGRESS)
1133 				goto finish;
1134 		}
1135 	}
1136 	return;
1137 
1138 finish:
1139 	/* finish curent request */
1140 	atmel_sha_finish_req(dd->req, err);
1141 }
1142 
1143 static irqreturn_t atmel_sha_irq(int irq, void *dev_id)
1144 {
1145 	struct atmel_sha_dev *sha_dd = dev_id;
1146 	u32 reg;
1147 
1148 	reg = atmel_sha_read(sha_dd, SHA_ISR);
1149 	if (reg & atmel_sha_read(sha_dd, SHA_IMR)) {
1150 		atmel_sha_write(sha_dd, SHA_IDR, reg);
1151 		if (SHA_FLAGS_BUSY & sha_dd->flags) {
1152 			sha_dd->flags |= SHA_FLAGS_OUTPUT_READY;
1153 			if (!(SHA_FLAGS_CPU & sha_dd->flags))
1154 				sha_dd->flags |= SHA_FLAGS_DMA_READY;
1155 			tasklet_schedule(&sha_dd->done_task);
1156 		} else {
1157 			dev_warn(sha_dd->dev, "SHA interrupt when no active requests.\n");
1158 		}
1159 		return IRQ_HANDLED;
1160 	}
1161 
1162 	return IRQ_NONE;
1163 }
1164 
1165 static void atmel_sha_unregister_algs(struct atmel_sha_dev *dd)
1166 {
1167 	int i;
1168 
1169 	for (i = 0; i < ARRAY_SIZE(sha_1_256_algs); i++)
1170 		crypto_unregister_ahash(&sha_1_256_algs[i]);
1171 
1172 	if (dd->caps.has_sha224)
1173 		crypto_unregister_ahash(&sha_224_alg);
1174 
1175 	if (dd->caps.has_sha_384_512) {
1176 		for (i = 0; i < ARRAY_SIZE(sha_384_512_algs); i++)
1177 			crypto_unregister_ahash(&sha_384_512_algs[i]);
1178 	}
1179 }
1180 
1181 static int atmel_sha_register_algs(struct atmel_sha_dev *dd)
1182 {
1183 	int err, i, j;
1184 
1185 	for (i = 0; i < ARRAY_SIZE(sha_1_256_algs); i++) {
1186 		err = crypto_register_ahash(&sha_1_256_algs[i]);
1187 		if (err)
1188 			goto err_sha_1_256_algs;
1189 	}
1190 
1191 	if (dd->caps.has_sha224) {
1192 		err = crypto_register_ahash(&sha_224_alg);
1193 		if (err)
1194 			goto err_sha_224_algs;
1195 	}
1196 
1197 	if (dd->caps.has_sha_384_512) {
1198 		for (i = 0; i < ARRAY_SIZE(sha_384_512_algs); i++) {
1199 			err = crypto_register_ahash(&sha_384_512_algs[i]);
1200 			if (err)
1201 				goto err_sha_384_512_algs;
1202 		}
1203 	}
1204 
1205 	return 0;
1206 
1207 err_sha_384_512_algs:
1208 	for (j = 0; j < i; j++)
1209 		crypto_unregister_ahash(&sha_384_512_algs[j]);
1210 	crypto_unregister_ahash(&sha_224_alg);
1211 err_sha_224_algs:
1212 	i = ARRAY_SIZE(sha_1_256_algs);
1213 err_sha_1_256_algs:
1214 	for (j = 0; j < i; j++)
1215 		crypto_unregister_ahash(&sha_1_256_algs[j]);
1216 
1217 	return err;
1218 }
1219 
1220 static bool atmel_sha_filter(struct dma_chan *chan, void *slave)
1221 {
1222 	struct at_dma_slave	*sl = slave;
1223 
1224 	if (sl && sl->dma_dev == chan->device->dev) {
1225 		chan->private = sl;
1226 		return true;
1227 	} else {
1228 		return false;
1229 	}
1230 }
1231 
1232 static int atmel_sha_dma_init(struct atmel_sha_dev *dd,
1233 				struct crypto_platform_data *pdata)
1234 {
1235 	int err = -ENOMEM;
1236 	dma_cap_mask_t mask_in;
1237 
1238 	/* Try to grab DMA channel */
1239 	dma_cap_zero(mask_in);
1240 	dma_cap_set(DMA_SLAVE, mask_in);
1241 
1242 	dd->dma_lch_in.chan = dma_request_slave_channel_compat(mask_in,
1243 			atmel_sha_filter, &pdata->dma_slave->rxdata, dd->dev, "tx");
1244 	if (!dd->dma_lch_in.chan) {
1245 		dev_warn(dd->dev, "no DMA channel available\n");
1246 		return err;
1247 	}
1248 
1249 	dd->dma_lch_in.dma_conf.direction = DMA_MEM_TO_DEV;
1250 	dd->dma_lch_in.dma_conf.dst_addr = dd->phys_base +
1251 		SHA_REG_DIN(0);
1252 	dd->dma_lch_in.dma_conf.src_maxburst = 1;
1253 	dd->dma_lch_in.dma_conf.src_addr_width =
1254 		DMA_SLAVE_BUSWIDTH_4_BYTES;
1255 	dd->dma_lch_in.dma_conf.dst_maxburst = 1;
1256 	dd->dma_lch_in.dma_conf.dst_addr_width =
1257 		DMA_SLAVE_BUSWIDTH_4_BYTES;
1258 	dd->dma_lch_in.dma_conf.device_fc = false;
1259 
1260 	return 0;
1261 }
1262 
1263 static void atmel_sha_dma_cleanup(struct atmel_sha_dev *dd)
1264 {
1265 	dma_release_channel(dd->dma_lch_in.chan);
1266 }
1267 
1268 static void atmel_sha_get_cap(struct atmel_sha_dev *dd)
1269 {
1270 
1271 	dd->caps.has_dma = 0;
1272 	dd->caps.has_dualbuff = 0;
1273 	dd->caps.has_sha224 = 0;
1274 	dd->caps.has_sha_384_512 = 0;
1275 
1276 	/* keep only major version number */
1277 	switch (dd->hw_version & 0xff0) {
1278 	case 0x420:
1279 		dd->caps.has_dma = 1;
1280 		dd->caps.has_dualbuff = 1;
1281 		dd->caps.has_sha224 = 1;
1282 		dd->caps.has_sha_384_512 = 1;
1283 		break;
1284 	case 0x410:
1285 		dd->caps.has_dma = 1;
1286 		dd->caps.has_dualbuff = 1;
1287 		dd->caps.has_sha224 = 1;
1288 		dd->caps.has_sha_384_512 = 1;
1289 		break;
1290 	case 0x400:
1291 		dd->caps.has_dma = 1;
1292 		dd->caps.has_dualbuff = 1;
1293 		dd->caps.has_sha224 = 1;
1294 		break;
1295 	case 0x320:
1296 		break;
1297 	default:
1298 		dev_warn(dd->dev,
1299 				"Unmanaged sha version, set minimum capabilities\n");
1300 		break;
1301 	}
1302 }
1303 
1304 #if defined(CONFIG_OF)
1305 static const struct of_device_id atmel_sha_dt_ids[] = {
1306 	{ .compatible = "atmel,at91sam9g46-sha" },
1307 	{ /* sentinel */ }
1308 };
1309 
1310 MODULE_DEVICE_TABLE(of, atmel_sha_dt_ids);
1311 
1312 static struct crypto_platform_data *atmel_sha_of_init(struct platform_device *pdev)
1313 {
1314 	struct device_node *np = pdev->dev.of_node;
1315 	struct crypto_platform_data *pdata;
1316 
1317 	if (!np) {
1318 		dev_err(&pdev->dev, "device node not found\n");
1319 		return ERR_PTR(-EINVAL);
1320 	}
1321 
1322 	pdata = devm_kzalloc(&pdev->dev, sizeof(*pdata), GFP_KERNEL);
1323 	if (!pdata) {
1324 		dev_err(&pdev->dev, "could not allocate memory for pdata\n");
1325 		return ERR_PTR(-ENOMEM);
1326 	}
1327 
1328 	pdata->dma_slave = devm_kzalloc(&pdev->dev,
1329 					sizeof(*(pdata->dma_slave)),
1330 					GFP_KERNEL);
1331 	if (!pdata->dma_slave) {
1332 		dev_err(&pdev->dev, "could not allocate memory for dma_slave\n");
1333 		return ERR_PTR(-ENOMEM);
1334 	}
1335 
1336 	return pdata;
1337 }
1338 #else /* CONFIG_OF */
1339 static inline struct crypto_platform_data *atmel_sha_of_init(struct platform_device *dev)
1340 {
1341 	return ERR_PTR(-EINVAL);
1342 }
1343 #endif
1344 
1345 static int atmel_sha_probe(struct platform_device *pdev)
1346 {
1347 	struct atmel_sha_dev *sha_dd;
1348 	struct crypto_platform_data	*pdata;
1349 	struct device *dev = &pdev->dev;
1350 	struct resource *sha_res;
1351 	int err;
1352 
1353 	sha_dd = devm_kzalloc(&pdev->dev, sizeof(*sha_dd), GFP_KERNEL);
1354 	if (sha_dd == NULL) {
1355 		dev_err(dev, "unable to alloc data struct.\n");
1356 		err = -ENOMEM;
1357 		goto sha_dd_err;
1358 	}
1359 
1360 	sha_dd->dev = dev;
1361 
1362 	platform_set_drvdata(pdev, sha_dd);
1363 
1364 	INIT_LIST_HEAD(&sha_dd->list);
1365 	spin_lock_init(&sha_dd->lock);
1366 
1367 	tasklet_init(&sha_dd->done_task, atmel_sha_done_task,
1368 					(unsigned long)sha_dd);
1369 
1370 	crypto_init_queue(&sha_dd->queue, ATMEL_SHA_QUEUE_LENGTH);
1371 
1372 	sha_dd->irq = -1;
1373 
1374 	/* Get the base address */
1375 	sha_res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
1376 	if (!sha_res) {
1377 		dev_err(dev, "no MEM resource info\n");
1378 		err = -ENODEV;
1379 		goto res_err;
1380 	}
1381 	sha_dd->phys_base = sha_res->start;
1382 
1383 	/* Get the IRQ */
1384 	sha_dd->irq = platform_get_irq(pdev,  0);
1385 	if (sha_dd->irq < 0) {
1386 		dev_err(dev, "no IRQ resource info\n");
1387 		err = sha_dd->irq;
1388 		goto res_err;
1389 	}
1390 
1391 	err = devm_request_irq(&pdev->dev, sha_dd->irq, atmel_sha_irq,
1392 			       IRQF_SHARED, "atmel-sha", sha_dd);
1393 	if (err) {
1394 		dev_err(dev, "unable to request sha irq.\n");
1395 		goto res_err;
1396 	}
1397 
1398 	/* Initializing the clock */
1399 	sha_dd->iclk = devm_clk_get(&pdev->dev, "sha_clk");
1400 	if (IS_ERR(sha_dd->iclk)) {
1401 		dev_err(dev, "clock initialization failed.\n");
1402 		err = PTR_ERR(sha_dd->iclk);
1403 		goto res_err;
1404 	}
1405 
1406 	sha_dd->io_base = devm_ioremap_resource(&pdev->dev, sha_res);
1407 	if (!sha_dd->io_base) {
1408 		dev_err(dev, "can't ioremap\n");
1409 		err = -ENOMEM;
1410 		goto res_err;
1411 	}
1412 
1413 	err = clk_prepare(sha_dd->iclk);
1414 	if (err)
1415 		goto res_err;
1416 
1417 	atmel_sha_hw_version_init(sha_dd);
1418 
1419 	atmel_sha_get_cap(sha_dd);
1420 
1421 	if (sha_dd->caps.has_dma) {
1422 		pdata = pdev->dev.platform_data;
1423 		if (!pdata) {
1424 			pdata = atmel_sha_of_init(pdev);
1425 			if (IS_ERR(pdata)) {
1426 				dev_err(&pdev->dev, "platform data not available\n");
1427 				err = PTR_ERR(pdata);
1428 				goto iclk_unprepare;
1429 			}
1430 		}
1431 		if (!pdata->dma_slave) {
1432 			err = -ENXIO;
1433 			goto iclk_unprepare;
1434 		}
1435 		err = atmel_sha_dma_init(sha_dd, pdata);
1436 		if (err)
1437 			goto err_sha_dma;
1438 
1439 		dev_info(dev, "using %s for DMA transfers\n",
1440 				dma_chan_name(sha_dd->dma_lch_in.chan));
1441 	}
1442 
1443 	spin_lock(&atmel_sha.lock);
1444 	list_add_tail(&sha_dd->list, &atmel_sha.dev_list);
1445 	spin_unlock(&atmel_sha.lock);
1446 
1447 	err = atmel_sha_register_algs(sha_dd);
1448 	if (err)
1449 		goto err_algs;
1450 
1451 	dev_info(dev, "Atmel SHA1/SHA256%s%s\n",
1452 			sha_dd->caps.has_sha224 ? "/SHA224" : "",
1453 			sha_dd->caps.has_sha_384_512 ? "/SHA384/SHA512" : "");
1454 
1455 	return 0;
1456 
1457 err_algs:
1458 	spin_lock(&atmel_sha.lock);
1459 	list_del(&sha_dd->list);
1460 	spin_unlock(&atmel_sha.lock);
1461 	if (sha_dd->caps.has_dma)
1462 		atmel_sha_dma_cleanup(sha_dd);
1463 err_sha_dma:
1464 iclk_unprepare:
1465 	clk_unprepare(sha_dd->iclk);
1466 res_err:
1467 	tasklet_kill(&sha_dd->done_task);
1468 sha_dd_err:
1469 	dev_err(dev, "initialization failed.\n");
1470 
1471 	return err;
1472 }
1473 
1474 static int atmel_sha_remove(struct platform_device *pdev)
1475 {
1476 	static struct atmel_sha_dev *sha_dd;
1477 
1478 	sha_dd = platform_get_drvdata(pdev);
1479 	if (!sha_dd)
1480 		return -ENODEV;
1481 	spin_lock(&atmel_sha.lock);
1482 	list_del(&sha_dd->list);
1483 	spin_unlock(&atmel_sha.lock);
1484 
1485 	atmel_sha_unregister_algs(sha_dd);
1486 
1487 	tasklet_kill(&sha_dd->done_task);
1488 
1489 	if (sha_dd->caps.has_dma)
1490 		atmel_sha_dma_cleanup(sha_dd);
1491 
1492 	clk_unprepare(sha_dd->iclk);
1493 
1494 	return 0;
1495 }
1496 
1497 static struct platform_driver atmel_sha_driver = {
1498 	.probe		= atmel_sha_probe,
1499 	.remove		= atmel_sha_remove,
1500 	.driver		= {
1501 		.name	= "atmel_sha",
1502 		.of_match_table	= of_match_ptr(atmel_sha_dt_ids),
1503 	},
1504 };
1505 
1506 module_platform_driver(atmel_sha_driver);
1507 
1508 MODULE_DESCRIPTION("Atmel SHA (1/256/224/384/512) hw acceleration support.");
1509 MODULE_LICENSE("GPL v2");
1510 MODULE_AUTHOR("Nicolas Royer - Eukréa Electromatique");
1511